Interview with Li Dongrong, Former Deputy Governor of the Central Bank: We Must Accelerate the Development of Safety Standards for Large Model Technology in Financial Applications to Avoid Algorithm Discrimination and Precise Exclusion Caused by Technology Distortion

Everyday Economic News Reporter | Zhang Rui   Editor | Dong Xing Sheng

On March 5th, the opening session of the Fourth Session of the 14th National People’s Congress was held at the Great Hall of the People in Beijing.

Premier Li Qiang delivered the government work report, outlining the main goals, strategic tasks, and major projects for the “14th Five-Year Plan” period.

Previously, the “14th Five-Year Plan” recommendations explicitly proposed accelerating the development of a strong financial nation; vigorously developing financial technology, green finance, inclusive finance, pension finance, and digital finance.

Against the backdrop of fully implementing the “Five Major Articles” of financial development, how should financial institutions avoid “digitizing for the sake of digitization”? Can the current regulatory framework keep pace with the rapid iteration of AI-driven finance? How can we prevent highly precise data profiling from excluding the most needy groups, such as flexible workers, from essential financial services?

With these questions in mind, during the National Two Sessions, Daily Economic News (NBD) interviewed Li Dongrong, former Vice Governor of the People’s Bank of China.

Since graduating from university over 40 years ago, Li Dongrong has been rooted in the financial sector, serving as Vice Governor of the Guangdong Branch of the People’s Bank of China, Deputy Director of the State Administration of Foreign Exchange, Assistant Governor and Vice Governor of the People’s Bank of China, and participating in five national financial work conferences since 1997. In June 2015, Li was responsible for establishing the China Internet Finance Association and served as its first president for seven years, witnessing the industry’s evolution from reckless growth and risk rectification to regulated development.

Digitalization of financial institutions must focus on problems and results

NBD: You have repeatedly mentioned that digital finance is the “link and booster” of the “Five Major Articles” of finance. In the current context of vigorously developing technology finance, green finance, inclusive finance, and pension finance, many worry that digital finance might become merely a “tool.” How do you think financial institutions should avoid “digitizing for the sake of digitization” and truly integrate digital technology into these four major areas to solve real problems, rather than just superficial process automation?

Li Dongrong: To address this concern, I believe the key is to always stay true to the purpose of financial services, and to correctly grasp the application of digital technology while following the basic laws of financial operation.

First, the concept must be clear. Promoting digital finance in financial institutions should adhere to a customer-centric philosophy—meaning proactively viewing the entire business process from the customer’s perspective, focusing on pain points, difficulties, and bottlenecks in customer transactions, deeply understanding the differentiated needs of various customer groups, and driving a deep shift from “product-centered” to “customer-centered” service models.

Online and intelligent features are manifestations of digital finance advancement. Their fundamental goal is to enhance customers’ sense of gain, satisfaction, and trust, empowering financial service quality and efficiency through digital transformation, and better serving the real economy and people’s better lives.

Second, it must be problem-oriented and result-driven, resolutely avoiding digitalization for its own sake. Only by targeting specific issues and solving them precisely can the quality and effectiveness of digital transformation be guaranteed. Currently, some financial institutions still face many problems in advancing digital finance—for example, lacking precise understanding of customer needs, leading to a disconnect between technological investment and actual demand; insufficient research, focusing more on technology development than practical application; systems that are difficult to adapt to business scenarios after launch, failing to effectively address core issues like low service efficiency and weak risk control.

Focusing on problems means addressing issues such as information asymmetry, unreasonable resource allocation, inaccurate risk identification, low service efficiency, and insufficient coverage of inclusive finance. Digital technology should be deeply integrated with these real problems to provide better solutions and avoid a “scattershot” approach to transformation.

Meanwhile, results orientation involves ensuring that digital initiatives aim to improve service quality, enhance core competitiveness, and achieve sustainable development. Only by combining problem-solving with result-driven strategies can financial institutions deepen their digital transformation, truly improve quality and efficiency, empower business growth, and solidify the foundation for digital development.

Finally, it is essential to uphold the bottom line of security and compliance. Innovation and application must match risk management capabilities, maintaining a risk bottom line. Strengthening customer privacy protection and risk prevention, strictly adhering to relevant laws and regulations on financial supervision, data security, and personal information protection, ensures that digital transformation proceeds safely, steadily, and lawfully.

Accelerate the formulation of security standards for large model technology applications in finance

NBD: You have repeatedly emphasized that “14th Five-Year Plan” period intelligent finance will become a key direction of digital finance, and that AI intelligent agents and industry large models are already penetrating core financial operations. Considering issues like limited interpretability and algorithmic bias, how should financial institutions, especially small and medium-sized ones, balance technological application and risk control, and find a development path suited to their capabilities? Can the current regulatory framework outpace the rapid iteration of AI in finance? Do regulators need to introduce specific governance frameworks for intelligent finance?

Li Dongrong: I will share my views on this from three aspects:

First, compared to large institutions, small and medium-sized financial institutions face significant resource gaps in applying intelligent finance—smaller data volumes and dimensions, weaker digital infrastructure, and insufficient talent reserves. However, they also have advantages such as flexible mechanisms, shorter decision chains, proximity to real economy entities, and focused scenarios. They must identify their positioning, leverage strengths, and pursue differentiated, specialized development paths aligned with their realities.

They should aim to be small but refined, focused and strong. Based on their resource endowments, they should concentrate on regional core scenarios like local industries, supply chain finance, rural revitalization, and similar areas. Using lightweight, scenario-based digital methods, they can develop specialized products, refine services, and optimize risk control, creating their own competitive “moats.”

Prioritize scenarios with high frequency and immediate impact. For example, in inclusive lending, micro risk control, counter operations, customer service, and compliance management, focus on areas with small investment, quick results, and manageable risks. Use intelligent applications to directly address practical issues.

Highlight regional and customer group characteristics. Incorporate local industry and micro-business data into big data models to improve risk control accuracy and service adaptability, forming distinctive scenario-based intelligent financial capabilities.

Second, open-source, high-performance, low-cost large model technologies like DeepSeek provide unprecedented opportunities for small and medium-sized institutions to accelerate AI application. However, the quality of data, compliance, and risk management are critical to ensuring safe and reliable AI operation.

Therefore, when advancing intelligent finance, small and medium institutions should pay special attention to:

1. Strengthening data governance. In the digital age, data governance is the core of competition—focused on proper data collection, deep mining, and full utilization. Small and medium institutions often have limited customer bases and data quality issues. They must improve data governance by ensuring data security, systematizing and standardizing data management, and building management platforms. Promoting data asset sharing and integration is also essential.

2. Enhancing risk prevention. Strengthen human-machine collaboration, leveraging AI’s strengths in data processing, risk identification, and modeling. Establish robust review and approval mechanisms, clarify responsibilities between humans and machines, and ensure human oversight in key steps and decisions to keep risks under control.

Third, it is necessary to improve the governance system for intelligent finance. Strengthen regulatory science capabilities, accelerate the development of standards for the safe application of large models in finance, and use these standards to guide compliant and secure development within a regulatory framework.

Providing warm, intelligent financial services with AI

NBD: In your recent speech, you mentioned that inclusive finance is shifting from “whether it exists” to “how good it is,” and highlighted care for new employment groups like couriers and ride-hail drivers. These groups have highly unstable employment and income, making traditional risk models difficult to apply. While using DeepSeek and other new technologies to reduce service costs, how can we avoid highly precise data profiling from “excluding” these most needy groups from essential financial services? Where is the balance?

Li Dongrong: First, in promoting digital finance, we adhere to the principle of “technology neutrality,” encouraging the reasonable application and innovation of new technologies. “Technology neutrality” means that technology itself is inherently fair, without good or evil, or bias. The ultimate impact and value of technology depend on how it is designed and applied—specifically, the scenarios and rules under which it operates.

Therefore, in practice, we should focus on regulating the application of technology to better serve those in need of financial services, rather than “labeling” and excluding them, which would contradict the core principles of inclusive, fair, and equitable finance.

Second, regulators should strengthen guidance on technological application behaviors, emphasizing human-centered and technology-for-good principles. Improving standards and norms for intelligent finance can prevent issues like algorithmic discrimination and exclusion caused by technological biases, making technology a tool to bridge financial gaps and improve service quality. Financial institutions should use technology to lower service barriers, not to set up data-driven walls. Regulatory sandboxes can help improve supervision efficiency, promote innovation, and balance efficiency with safety.

Finally, financial institutions must stay true to their original mission of serving the real economy and the public, continuously innovate products and services, and actively use AI and big data models to assess creditworthiness and provide warm, intelligent financial services. They should strictly adhere to legal and ethical standards, implementing AI ethics governance and training, and improving measures in data collection, algorithm design, product development, and application to avoid biases related to ethnicity, religion, gender, age, occupation, and other factors.

For example, to serve flexible employment groups like couriers and ride-hail drivers, financial institutions can strengthen cooperation with relevant platforms, legally and compliantly analyze behavioral data such as order stability and account transaction patterns, and build differentiated credit evaluation models. These can be integrated into trustworthy credit profiles, enabling “behavior-based credit scoring and credit limit granting.”

In summary, applying intelligent finance must uphold the political and people-centered nature of financial services. Through regulations and standards, we can guide the ethical and responsible use of AI technology to better meet the financial needs of the people.